CASC15 mediated upregulation of SERPINE1 by sponging miR-30c- 5p correlates with poor prognosis and immune inltration of gastric cancer

Background: As a major regulator of plasminogen activator system (PAs), SERPINE1(PAI-1) is associated with poor prognosis in a variety of cancers, but in Stomach adenocarcinoma (STAD), its mechanism of action is not so clear. Methods(cid:0)Based on RNA Sequencing dataset and survival data for TCGA-STAD, this study screened out genes of signicant difference between the gastric cancer group and the normal group using the R language limma package, and carried out a pan-cancer analysis on their expression in combination with the Genotype-Tissue Expression (GTEx) database at the same time. Subsequently, the miRNA and lncRNA associated with high expression level of PAI-1 signicantly were identied through the Starbase database in combination with a series of analyses on expression level, correlation and survival. The correlation of immune cell inltration, biomarkers of immune cells and immune checkpoints between PAI-1 were investigated via TIMER and GEPIA database. the of STAD patients. According to analysis results, CASC15/hsa-miR-30c-5p/ PAI-1 may be the most potential regulatory network of PAI-1 related to survival in gastric cancer. And the expression level of PAI-1 in gastric cancer had a positive correlation with immune cell inltration, biomarkers of immune cells and immune checkpoints obviously. and


Introduction
Globally, as the fth most common malignancy, gastric cancer is also the third most important cancercausing death. The high-risk factors for gastric cancer include helicobacter pylori infection, poor diet, age and high salt intake. (1)The combination of surgery, chemotherapy and targeted therapy has prolonged the overall survival of patients with gastric cancer signi cantly, and improved their quality of life as well.
However, most of the patients with gastric cancer have been at an advanced stage with poor prognosis once con rmed as the disease for lack of obvious or speci c symptoms in an early stage. (2) Therefore, it is an urgent to nd new therapeutic targets or prognostic biomarkers.
As a serine protease inhibitor, SERPINE1(also known as plasminogen activator inhibitor type 1[PAI-1]), is a member of the urokinase plasminogen activator (uPA) system(3)that regulates extracellular brinolysis and in uences cell invasion and migration, as well as ECM remodeling.(4)PAI-1 is closely related to the occurrence, development, invasion and metastasis of various cancers. PAI-1 induces migration and invasion of esophageal squamous cell carcinoma (ESCC) cells and macrophages by low-density lipoprotein receptor-related protein 1 (LRP1) of PAI-1 receptor based on Akt and ERK1/2 signaling pathways, thus promoting the progression of ESCC.(5) On the one hand, it is associated with poor prognosis in patients with different types of tumors, including breast cancer.(6) On the other hand, the expression of PAI-1 may promote angiogenesis and migration and anti-apoptosis, thus regulating the proliferation of cancer cells and support the tumorigenesis. (7) Similarly, some study has also identi ed PAI-1 in pancreatic tumor, which is closely related to venous thromboembolism for pancreatic cancer. (8) A study has also suggested that in an oxygen-de cient environment, PAI-1 can induce a malignant phenotype of tumor as hypoxia-inducible factor through activating or enhancing JAK-STAT signaling pathway, TGF-β signaling pathway and NOTCH signaling pathway. (9)However, the prognosis, expression and relevant mechanism of PAI-1 in gastric cancer, as well as its relation with immune in ltration of gastric cancer, are still to be further investigated.
In this study, the TCGA-STAD dataset and GTEx dataset were utilized to analyze the expression, prognosis and survival of PAI-1 in gastric cancer, as well as its expression in various cancers. At the same time, the Starbase database was used to analyze PAI-1 related lncRNA and miRNA based on the sponge regulation mechanism of ceRNA, and thus, the lncRNA/miRNA/mRNA axis was determined. In addition, the study found that PAI-1 was positively correlated with immune cell in ltration, immune checkpoints, and biomarkers of immune cells in gastric cancer. The results suggested that PAI-1 regulated by lncRNA through ceRNA mediation may be a new biomarker for the treatment of gastric cancer.

Methods
Data download, processing and differential analysis The genomic and clinicopathological information of 33 cancers were obtained from The Cancer Genome Atlas (TCGA) database (https://genome-cancer.ucsc.edu/), and 18 tumors (BLCA BRCA CHOL COAD ESCA GBM HNSC KICH KIRC KIRP LIHC LUAD LUSC PRAD READ STAD THCA and UCEC) types with normal samples larger than 5 were further analyzed. A differential analysis of PAI-1 in the above cancers was performed using the R Package limma.(10) If P < 0.05, there was statistical signi cance. At the same time, differentially expressed genes DEGs in gastric cancer tissues and normal tissues were screened out by using R Package limma. The critical condition for screening was |log2FC| ≥ 1.0 and adj. P < 0.05, with p value corrected by the false discovery rate (FDR) method. Then, the survival package in R was used for a prognostic survival analysis on signi cant DEGs in gastric cancer. p-value < 0.05 were considered signi cant.

Analysis on GEPIA database
As a database, GEPIA (http://gepia.cancer-pku.cn/) is a web-based tool for cancer or normal geneexpression pro ling and interaction analysis based on TCGA and genotype-tissue expression (GTEx) data. (11)The expression of PAI-1 in 18 different cancers and normal tissues was investigated using GEPIA database. p-value < 0.05 were considered signi cant. Furthermore, 11 cancers types were analyzed for survival outcomes (including OS and RFS). In addition, GEPIA was also used for a survival analysis on upstream lncRNAs of PAI-1. Log rank p value <0.05 was considered as statistically signi cant. Finally, GEPIA was also used for evaluating the correlation of PAI-1 with the biomarkers of immune cells and immune checkpoints (CD274, CTLA4 and PDCD1). P < 0.05 and |R| > 0.1 as the thresholds, and if relevant conditions were met, there was statistical signi cance.

Prediction of candidate miRNA by Starbase database
Upstream miRNAs that may bind to PAI-1 were predicted by 7 prediction programs of target genes embedded in Starbase 2.0(https://rna.sysu.edu.cn/) online site (12), i.e., PitA, RNA22, miRmap, microT, Miranda, PicTar and TargetScan. Only miRNAs obtained by 3 or more target gene prediction programs can be used as candidate miRNAs.
Screening candidate miRNA and constructing miRNA -PAI-1 correlation analysis The expression data of miRNA in gastric cancer were downloaded from TCGA database. The R language limma package was used to screen out candidate miRNAs different in expression in gastric cancer tissues and normal tissues. p-value < 0.05 were considered statistically signi cant. Then, the spearman correlation coe cient method(13) was used to analyze the correlation between candidate miRNA and PAI-1, with P < 0.05 and |R| > 0.1 as the thresholds for screening. R language ggpubr package and ggExtra package were used to map the correlation between the candidate miRNA and PAI-1 expression level. Finally, Cytoscape software (V.3.8.0) was used to visualize the miRNA-PAI-1 regulatory network conforming to the above screening conditions. (14) Kaplan and Meier plotter analysis Kaplan and Meier plotter database (http://kmplot.com/analysis/) may be used to assess the effects of genes or miRNAs on survival in more than 20 cancers, including liver cancer. (15) The database was used for a survival analysis on hsa-miR-30c-5p in gastric cancer. For logrank. Log rank p-value < 0.05 were considered statistically signi cant.

LncRNA prediction and correlation analysis
The upstream lncRNAs of miR-30c-5p were predicted by miRNA-lncRNA module group of Starbase database. Screened lncRNAs were visualized by using Cytoscape software. Differentially expressed lncRNAs between gastric cancer tissues and normal tissues were obtained by a difference analysis of candidate lncRNAs using R language Limma package. p-value < 0.05 were considered statistically signi cant. Then, the spearman correlation coe cient method was used for a correlation analysis on differentiated lncRNAs and miR-30c-5p. P < 0.05 and |R| > 0.1 were considered statistically signi cant. Finally, the lncRNAs that had been screened out were used for a correlation analysis with PAI-1. In addition, the expression boxplots of identi ed lncRNAs in gastric cancer tissues and normal tissues were charted out by utilizing ggboxplot package.
LncRNA survival analysis and lncRNA-miRNA-mRNA network construction In combination with clinical data of TCGA-STAD, a survival analysis on lncRNAs that had been screened out was performed using R language survival and survminer packages. At last, the GEPIA database was used for another survival analysis on the obtained lncRNAs, so as to further screen prognostic lncRNAs.
Immune cell in ltration analysis TIMER (https://cistrome.shinyapps.io/timer/) Database is a web server for comprehensive analysis of tumor-in ltrating immune cells.(16) TIMER software was used to analyze the correlation of PAI-1 expression level in gastric cancer with immune cell in ltration level and immune checkpoint expression level. p-value < 0.05 were considered statistically signi cant. Then, the spearman correlation coe cient method was used to analyze the correlation between PAI-1 and biomarkers of immune cells. To evaluate the difference in in ltration of 22 types of immune cells between low expression group and high expression group of PAI-1, CiberSort (https://cibersort.stanford.edu/about.php) was used for estimate immune cells proportion in tumor tissues accurately.(17)p-value < 0.05 were considered statistically signi cant. At the same time, the spearman correlation coe cient method was used to analyze the correlation between PAI-1 expression level and the in ltration level of 22 types of immune cells. If P < 0.05 and |R| > 0.1, there was correlation. Venn diagram was then used for acquiring the intersection, and thus, immune cells highly correlated with PAI-1 expression level were obtained.

An analysis on the expression of PAI-1 in pan-cancer
To evaluate the potential function and expression of PAI-1 in a variety of cancers, the study analyzed the expression of PAI-1 rst, as shown in Fig.1. The expression level of PAI-1 in BRCA, COAD, ESCA, GBM, HNSC, KIRC, READ, STAD or THCA was signi cantly higher than in normal tissue. However, the level was signi cantly down-regulated in CHOL, KICH, KIRP, or LIHC. There was no signi cant difference for the level in BLCA, LUAD, LUSC, PRAD and UCEC. Similarly, the mRNA expression data of gastric cancer were analyzed, and through a combination of Kaplan-Meier analysis and COX regression model, 82 survivalrelated genes highly correlated with prognosis of gastric cancer were screened out (Supplementary Table  S1), including high-risk PAI-1. Subsequently, the expression of PAI-1 in various cancer tissues was veri ed by the GEPIA database. As shown in Fig.2A-2H, the expression was up-regulated obviously in BRCA, COAD, ESCA, GBM, HNSC, KIRC, READ, and STAD, while down-regulated in KICH, KIRP, LIHC, LUSC, THCA, and UCEA ( Fig.2I-2N). Overall, the expression levels of PAI-1 in BRCA, COAD, ESCA, GBM, HNSC, KIRC, READ, and STAD obviously increased, while in KICH, KIRP, and LIHC decreased, indicating that PAI-1 might be closely related to the occurrence, development, invasion and metastasis of these 11 cancers.

An analysis on the prognosis of PAI-1 in various cancers
The survival analysis for PAI-1 in these 11 cancers was conducted based on the GEPIA online database.
As shown in Fig.3 and Fig.4, in terms of overall survival (OS), the high expression levels of PAI-1 in HNSC, STAD, KIRP and LIHC indicated a poor prognosis, while in terms of recurrence-free survival (RFS), the up-regulation of PAI-1 expression in GBM, KIRC, STAD, KICH and KIRP represented a poor prognosis. In combination with the expression levels of PAI-1 in cancer tissues and normal tissues, the study found that only the up-regulation of PAI-1 expression level worsened the prognosis of gastric cancer, and there was no statistical signi cance for the survival prediction related to other malignant tumors. To sum up, as a potential prognostic biomarker, PAI-1 may indicate poor prognosis of gastric cancer.
Prediction and analysis on upstream miRNAs of PAI-1 The upstream miRNAs of PAI-1 were predicted based on the Starbase online database, and the miRNAs that could be predicted by three databases simultaneously were captured for further analysis. According to the mechanism of action of miRNA on target gene mRNA, miRNA should be negatively correlated with target gene. Subsequently, the correlation between PAI-1 and these miRNAs was calculated, and at last, Prediction and analysis on the upstream lncRNAs of miR-30c-5p According to the competitive endogenous RNA hypothesis, lncRNA may be involved in the expression and regulation of target genes through competitive binding with miRNA. Through miRNA, the lncRNA that can bind to it may be derived reversely based on this mechanism. Thus, a total of 101 lncRNAs that may bind to miR-30c-5p were obtained through a prediction based on the Starbase database, and the lncRNA-miR-30c-5p regulatory network was constructed using Cytoscape software ( Supplementary Fig.1). A screening based on the mechanism of competitive binding with miRNA identi ed 6 lncRNAs(GATA3-AS1, NORAD, KCNH7-AS1, LINC02863, CASC15, and LINC01094) expression up-regulated signi cantly in gastric cancer as compared with those in the normal group and signi cantly correlated with miR-30c-5p (P < 0.05, |R| > 0.1). For their expression levels, shown Fig.6A-6F. Subsequently, the 6 lncRNAs were performed correlation analysis with PAI-1, and the results were as shown in Table 2. Obviously, the p value for correlation of GATA3-AS1, NORAD, and KCNH7-AS1 with PAI-1 was greater than 0.05, NORAD was negatively correlated with PAI-1, and therefore, there was no statistical signi cance. Next, LINC02863, CASC15 and LINC01094 were analyzed for survival, and the results showed that in gastric cancer with high expression levels of CASC15 and LINC01094, the OS was poor. Finally, again with the GEPIA database, the prognosis of CASC15 and LINC01094 in gastric cancer was veri ed. As shown in Figure 7, only patients with high expression level of CASC15 in gastric cancer had poor OS and RFS, indicating that the up-regulation of CASC15 expression level was related to the poor prognosis of gastric cancer. Therefore, CASC15 may be the most promising upstream lncRNA of miR-30c-5p/ PAI-1.

PAI-1 positively correlates with immune cell in ltration in gastric cancer
The copy number variation (CNV) of gene, a variant form of DNA mutation, has been reported to be closely related to human tumors.The study analyzed the relations of various copy numbers of PAI-1 with B cell, CD8+T cell, CD4+T cell, macrophage, neutrophil, or dendritic cell in ltration degree using TIMER database. The CNV degree was indicated by Deep Deletion, Arm−level Deletion, Diploid/Normal, Arm−level Gain and high Ampli cation. As shown in Fig.8, in gastric cancer with CNV of PAI-1, except for deep deletion, the in ltration degree of most immune cells was signi cantly reduced, indicating that the PAI-1 might mediated immune cells in ltration. Also, the correlation between PAI-1 and the in ltration degree of the 6 types of immune cells was veri ed through tumor purity correction. As shown in Fig.9B and Fig.9D-9F, the expression level of PAI-1 had a signi cant positive correlation with the in ltration degree of CD8+T cell, Macrophage, Neutrophil, and Dendritic cell. As shown in Fig.9A and Fig.9C, the correlation between B cell, CD4+T cell and expression level of PAI-1 was no statistical signi cance. Therefore, CiberSort was further used to analyze the proportion of each of the 22 types of immune cells in all samples (Fig.10). Then, PAI-1 was divided into high and low expression groups according to the corresponding expression levels, and the violin diagram was used to observe the difference in in ltration degree of various immune cells for different PAI-1 expression levels. The results showed that the in ltration degrees of NK cells resting, Monocytes, Dendritic cells activated, Mast cells activated, Eosinophils, and Neutrophils in the high PAI-1 expression group were signi cantly increased as compared with those in the low PAI-1 expression group, and the in ltration degree of NK cells activated in the high PAI-1 expression group was signi cantly decreased as compared with that in the low PAI-1 expression group (Fig.11A). Furthermore, as shown in Fig.11B, in the analysis concerning correlation between PAI-1 expression level and immune cell in ltration degree, the PAI-1 expression level had a signi cant positive correlation with the in ltration degrees of NK cells resting, Monocytes, Dendritic cells activated, Mast cells activated, Eosinophils, and Neutrophils, and a negative correlation with the in ltration degree of NK cells activated. Two methods indicated that the in ltration degrees of these seven immune cells were correlated with PAI-1.

Correlation between PAI-1 expression level and biomarkers of immune cells in gastric cancer
To further explore the immunization of PAI-1 in gastric cancer, the Spearman's rank correlation coe cient method was used for analyzing the correlation between PAI-1 and biomarkers of immune cells. As shown in Table 3  Association of PAI-1 with common immune checkpoints in gastric cancer tissues The commonly used immune checkpoints, including PD1, CD274 and CTLA4, are important checkpoints in the tumor immune escape mechanism. Based on the above analyses, it can be preliminarily assumed that PAI-1 has a potential for guiding the occurrence and development of gastric cancer. Therefore, the correlation of PAI-1 with PD1, CD274 and CTLA4 was analyzed using TIMER after tumor purity correction. The results showed that PAI-1 had a signi cant positive correlation with PD1, CD274 and CTLA4 (Fig.12A-12C). The validation of GEPIA database has also con rmed the above results ( Figure 12D-12F). The results have supported the involvement of tumor immune escape mechanism in gastric cancer development regulated by PAI-1.

Discussion
By now, the morbidity and mortality of gastric cancer have always been at a high level with poor prognosis. For this, it is urgent to nd new solutions. Therefore, the study on the molecular mechanism of gastric cancer is conducive to the search for potential biomarkers and the development of new therapeutic targets. At present, with the development of bioinformatics, the bioinformatics analysis method has been widely used to screen and analyze the genes related to the progression and prognosis of various cancers, and has provided new ideas for improving the studies on tumor molecular mechanism and exploring new targeted therapies.
More and more studies have shown that PAI-1 plays an important role in many cancers. (7,(18)(19)(20) Studies have shown that the high expression of PAI-1 in gastric cancer may promote the occurrence and development of gastric cancer, and at the same time, it is associated with the poor prognosis of gastric cancer. (21,22)However, the speci c mechanism of PAI-1 in the occurrence and development of gastric cancer is still to be further clari ed. Therefore, this study carried out a pan-cancer analysis of PAI-1 using TCGA database rstly, and identi ed highly expressed PAI-1 in a variety of cancers. Based on the GEPIA database, a survival analysis was conducted for the cancers with high PAI-1 expression levels, and the results showed that the high PAI-1 expression was correlated with the poor prognosis of patients with tumors obviously, especially in those with gastric cancer, the OS and RFS were signi cantly decreased among the patients with high PAI-1 expression.
According to relevant reports, non-coding genes including lncRNAs, miRNAs and circRNAs, may be involved in the in the regulation of mRNA through the ceRNA regulation mechanism, thus playing an important role in a variety of cancers. (23)(24)(25)The microRNAs are a category of small non-coding RNAs of approximate length 22 nucleotides that degrade mRNAs or inhibit mRNAs expression mainly through binding to miRNA response element (MRE) on target RNA transcript. Under some conditions, miRNA may occasionally enhance gene expression or increase the target gene expression level.(26) In order to identify miRNAs that may bind to PAI-1, 7 different versions of miRNA software were used for relevant prediction, and after screening based on correlation, 10 miRNAs were obtained at last. Theoretically, due to the upregulation of PAI-1 expression level, the expression levels of miRNAs bound to PAI-1 will be downregulated. Therefore, a further survival analysis was conducted using miR-30c-5p. The results showed that the increased expression level of miR-30c-5p in gastric cancer was correlated with the poor prognosis of gastric cancer signi cantly. Studies have shown that miR-30c-5p can inhibit cell growth and proliferation through targeting RAB32, and reducing the expression level of miR-30c-5p may promote the development of liver cancer. (27) Some studies have shown that the expression level of miR-30c-5p in gastric cancer is signi cantly reduced, and its downregulation may advance the capacities of gastric cancer cells for migration and invasion. Further studies have shown that through inhibiting the expression of its target transfer-related protein 1 (MTA1), miR-30c-5p can inhibit epithelial-to-mesenchymal transformation (EMT), an important process in gastric cancer metastasis.(28) Therefore, miR-30c-5p was selected as one of the most promising upstream targets of PAI-1.
Based on the hypothesis of ceRNA regulation, lncRNA can be used as endogenous RNA to adsorb miRNA, thus participating in the regulation of cancer tissues. (29) Therefore, at the upstream of miR-30c-5p/PAI-1 axis there should be carcinogenic lncRNA. Six signi cantly correlated lncRNAs that may bind to miR-30c-5p were identi ed through a prediction based on Starbase database. Then, through correlation analysis between lncRNAs and PAI-1, 3 candidate lncRNAs were screened out for survival analysis. At last, the most promising upstream lncRNA-CASC15 was obtained through a veri cation based on GEPIA database. Several studies have shown that CASC15 is highly expressed in liver cancer, lung cancer, tongue squamous carcinoma, gastric cancer, colorectal cancer, cervical cancer and breast cancer with carcinogenicity. (30)According to relevant reports, in gastric cancer, CASC15 is involved in the occurrence of gastric cancer when it interacts with EZH2 and WDR5 to regulate CDKN1A in the nucleus. Based on this molecular axis, cell apoptosis may be induced to promote or inhibit cell proliferation through upregulation or down-regulation of CASC15, and cell migration and invasion may be enhanced or inhibited through in uencing the EMT process. (31) In summary, CASC15/ miR-30c-5p/ PAI-1 axis may be a potential regulatory network in gastric cancer.
Numerous studies have proved that tumor immune cell in ltration may in uence the effect of chemotherapy, radiotherapy or immunotherapy and the prognosis of patients with tumors.(32, 33) Based on the TIMER database, the study has found that PAI-1 has a signi cant positive correlation with CD8+T cells, Macrophages, Neutrophils and Dendritic cells in gastric cancer. Moreover, PAI-1 is also positively correlated with most of the biomarkers of immune cells. These results have suggested that PAI-1 expression is closely related to immune cell in ltration in tumors. This may partially explain the carcinogenesis of the PAI-1.
Immunotherapy is not only dependent on the proportion of immune cell in ltration in tumor microenvironment, but also associated with the expression of tumor immune checkpoint.(34-36) Immune checkpoint is important for tumor immune escape mechanism.(37, 38)Therefore, the relationship between immune checkpoint and PAI-1 in gastric cancer was also analyzed. The results showed that the high expression of PAI-1 had a signi cant positive correlation with PD1, CD274 and CTLA4. Therefore, targeting PAI-1 may improve the effect of immunotherapy for gastric cancer indirectly.
In our study, the high expression of PAI-1 in a variety of human cancers was veri ed, and several analyses showed that the high expression of PAI-1 was correlated with poor prognosis of gastric cancer signi cantly. Also, the upstream miRNA and lncRNA of PAI-1 were predicted, and CASC15/ miR-30c-5p/ PAI-1 regulatory network was constructed. In addition, the relationship between PAI-1 and immune cell in ltration in gastric cancer was explored, and the results suggested that PAI-1 mediated the occurrence or progression of gastric cancer by changing tumor microenvironment and enhancing tumor immune escape mechanism.

Figure 1
The expression of SERPINE1 in 18 types of human cancer based on TCGA cancer and normal data. *p value < 0.05; **p value < 0.01; ***p value < 0.001.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.